Skip to content

fix(probe): print one refused-probe contract on the CLI and MCP plan - #1178

Merged
Birfy merged 6 commits into
mainfrom
agent/1175-probe-contract
Nov 20, 2026
Merged

Birfy merged 6 commits into
mainfrom
agent/1175-probe-contract

Conversation

@HugoXK

@HugoXK HugoXK commented Nov 20, 2026 •

Copy link
Copy Markdown
Collaborator

Fixes #1175.

Claimed under AGENTS.md "Claim, PR, release"; agent-ready is triage-side and applied to 0/18 open issues (repo-wide), so it was not waited on.

What changed

A refused probe printed two contracts: plan --probe --json (and algorithms run --probe --json) returned ok: true, ready: false, exit 1, while the MCP plan tool returned ok: false with no ready. A host that ported its check between surfaces read a dead subscription as ready, or a probe refusal as an invalid spec.

Both surfaces now print the same refusal: ok: false, ready: false, and next as the repair action. The CLI exit code stays 1 (the config answer, not the spec error that exits 2); the MCP refusal still never says start. The CLI ok flip is user-visible and carries changelog.d/1175-mcp-probe-contract.md; plain plan is unchanged and has no ready key on either surface. No model or network call in the suite -- the tests use a local 400 InvalidSubscription stub.

Contract decision: docs/algorithms.md (generated by tools/gen_algorithms_docs.py) documented ready: false; docs/plugin-quickstart.md said ok was about the plan itself; docs/plugins.md said the probe is the same on both surfaces; the MCP tool description said a refusal makes ok: false. The stricter union holds: a refusal is not success on either surface, so both keys are false everywhere. Docs, the shared SKILL.md and the rendered host copies (Claude Code, dsh plugin) say the same thing.

The release v0.11.0 shipped the old wording in CHANGELOG.md. The fragment now quotes that sentence verbatim and then states the corrected behaviour, so a reader of the release notes is not left assuming the two agree: it says a refused --json probe sets ready: false while leaving ok about the plan itself, and this PR sets ok: false there too.

Payloads at the tip (400 InvalidSubscription stub, one probe call per surface):

  • agentdescent plan spec.json --probe --json -> exit 1
    {"ok": false, "ready": false, "next": "renew the subscription at the provider, or point api_key_env / model at another plan"}
  • the plan tool over agentdescent mcp, {"spec": ..., "probe": true}
    {"ok": false, "ready": false, "next": "renew the subscription at the provider, or point api_key_env / model at another plan"}
  • algorithms run ... --probe --json -> exit 1, the same triple.
  • with a 200 stub: CLI exit 0 with ok: true, ready: true; MCP ok: true, ready: true.

Red test on 6b034fd2 (tests/test_mcp_plan_probe.py):

  • test_both_surfaces_publish_the_same_refusal_triple -> KeyError: 'ready' at tests/test_mcp_plan_probe.py:179
  • test_both_surfaces_publish_ok_and_ready_true_for_a_healthy_probe -> KeyError: 'ready' at tests/test_mcp_plan_probe.py:198
  • tests/test_plan_probe.py::test_plan_json_marks_a_refused_probe_unready_and_not_ok -> assert True is False at tests/test_plan_probe.py:360
  • tests/test_plan_probe.py::test_algorithms_run_json_exits_nonzero_on_a_refused_probe -> assert True is False at tests/test_plan_probe.py:251

Red-first re-verified at the new base 38f9d90f (the 0.11.0 release) with only the two test files taken from this branch: the same four ids fail the same way -- KeyError: 'ready' at tests/test_mcp_plan_probe.py:179 and :198, assert True is False at tests/test_plan_probe.py:360 and :251.

Tip SHA: a62bcf6

Ran:

  • pytest -q at c6000830 -> 22 failed, 10 errors, 7885 passed, 51 skipped (7968 collected). All 31 reproducible ids are red on pristine 14ba9529 too (docker, exec-bit, wheel-smoke environment); the 32nd, test_dsh_superagent_run.py::...[tmux], passes 2/2 in isolation at both tips and is full-suite-load flaky. None touches the probe, CLI plan, or MCP plan.
  • focused files at a62bcf65 (test_plan_probe, test_mcp_plan_probe, test_cli_help_epilog, test_cli_help_hides_internal_run, test_call_budget, test_budget_guard, test_integrations, test_plugin_host_cli) -> 191 passed, 2 skipped
  • python -m tools.gen_api_docs --check -> pass
  • python -m tools.gen_algorithms_docs --check -> pass
  • python -m tools.changelog_fragments check -> pass (1 fragment; the 0.11.0 release folded the rest away)
  • mkdocs build --strict -> pass (local config minus the social og-image plugin, which blocks on a font fetch in this sandbox; CI runs the full config). That stripped config was a temporary untracked file for the local run only: nothing in the dsh family (dsh-agent-loop, dsh-token-meter, dsh-session-*) covers a MkDocs build, so no hand-rolled capability ships from this branch.
    Not run:
  • full pytest -q after the merge -> the merge delta is the release fold plus dropping one changelog file; the focused files and the gates above were re-run at the tip. The full-suite count above is from c6000830, which carries the same code as this tip.
  • wheel smoke -> no wheel or release path change
  • real-model probe -> ARK_API_KEY is not in this box's environment; the stub is the measured 400 shape

No UI surface: this changes terminal/JSON payloads and an MCP tool description; there is no UI to screenshot.

main 38f9d90f (release v0.11.0) is merged in (fa5fe5b5, two parents). One conflict only: changelog.d/1093-probe-failed-next.md was modify/delete because the release folded it into CHANGELOG.md; the deletion is taken and this branch's edit to that file is dropped, since the contract now lives in changelog.d/1175-mcp-probe-contract.md. docs/plugin-quickstart.md and docs/plugins.md auto-merged, keeping main's 0.11.0 pins and this branch's text.

…1175)

A refused probe printed `ok: true` + `ready: false` on the CLI and
`ok: false` with no `ready` over MCP, so a host porting its check from one
surface to the other read a dead subscription as ready, or a probe refusal
as an invalid spec. Both surfaces now print `ok: false`, `ready: false`
and `next` as the repair; the CLI exit code stays 1.
@HugoXK HugoXK added agent-pr Pull request opened by an agent area:cli The agentdescent command line (cli.py, doctor, plan/evolve/status/apply) area:plugins Host plugins and installers: Claude Code, Codex, DSH, OpenCode, MCP server bug Something isn't working labels Nov 20, 2026
…1175)

The plugin embeds the SKILL.md body, so the probe paragraph it carries is
rendered from the package; regenerating keeps the checked-in copy from
drifting after the contract text changed.
@HugoXK
HugoXK marked this pull request as ready for review November 20, 2026 05:15
@HugoXK

HugoXK commented Nov 20, 2026

Copy link
Copy Markdown
Collaborator Author

Verdict: PASS (advisory — .github/agent/review-policy.json lists only Birfy, so this account records no counting PASS)

Tip SHA: 3b41def36b957bbcc08e845889326183f2e4b25f (read from the API in this turn; head matches the PR, non-draft, base 14ba9529, 4 commits, 14 files +139/−35). This is the first report on this PR.

Delta read: git diff c6000830..3b41def3 → 1 commit, 1 file, changelog.d/1093-probe-failed-next.md +2/−2. No code, test or doc path moved, so the review below is the code read at c6000830 applied to a byte-identical tree.

Ran (at this tip):

  • python -m tools.changelog_fragments check → 86 changelog fragment(s) OK, rc 0
  • python -m tools.gen_api_docs --check → docs/api.md is up to date; python -m tools.gen_algorithms_docs --check → docs/algorithms.md is up to date
  • pytest tests/test_plan_probe.py tests/test_mcp_plan_probe.py tests/test_cli_help_epilog.py tests/test_cli_help_hides_internal_run.py tests/test_budget.py tests/test_budget_guard.py tests/test_budget_governor.py tests/test_call_budget.py → 114 passed
  • pytest tests/test_integrations.py -k "checked_in_claude_plugin_matches_the_package or checked_in_dsh_plugin_matches_the_package" → 2 passed
  • repo-wide grep for about the plan itself → no hit in changelog.d/, docs/ or the skill copies
  • .github/scripts/risk_gate.py --pr 1178 → T3 (174 changed lines / 14 files)
  • behaviour measurements below were taken at c6000830; the code is identical at this tip (one text file in the delta)

Not run:

  • full pytest -q — it stalled on this host earlier and was killed, so I did not complete it
  • mkdocs build --strict — no mkdocs in my venv; the delta is a changelog fragment
  • CI — I did not wait for the test legs; this verdict rests on the local runs above

Summary, against the four questions:

  1. ready appears only on the probe branch. _with_probe (agentdescent/mcp.py:415-424) adds ready inside the probe call, and Tools.plan returns early for a plain plan and for spec errors, so a plain plan has no ready key; cli.py:2585 sets it only inside if probe is not None, and the only other writers are the --baseline-sample branches (:2627, :2655). The new no-probe test passes.
  2. Exit codes did not move. A spec error returns 2 with and without --probe; a refused probe returns 1 through the probe_refused guard (cli.py:2669: if not payload.get("ok", True) and not probe_refused: return 2, then return 0 if probe is None or probe.get("ok", True) else 1). _cmd_algorithms_run flips obj["ok"] after capturing ok, so the stage/pre return 2 is untouched and the JSON branch returns 1 via the probe check (:4867-4878). _print_plan never reads payload["ok"].
  3. The host copies are generated, not hand-written. test_checked_in_claude_plugin_matches_the_package and test_checked_in_dsh_plugin_matches_the_package pass at this tip, and both generated docs report up to date, so CLAUDE/dsh SKILL and docs/ carry the same contract text.
  4. Red-first ids reproduce at the base. On 14ba9529 with only the tip's test files: KeyError: 'ready' at tests/test_mcp_plan_probe.py:179 and :198, assert True is False at tests/test_plan_probe.py:360 and :251. The test files are unchanged in this delta.

The delta itself is correct: the fragment no longer says --json "leav[es] ok about the plan itself", which this PR makes false; it now says only `--json` sets `ready: false`, true in both the plan --probe and algorithms run --probe refusal branches, and it is silent on ok — which the concurrently-folded changelog.d/1175-mcp-probe-contract.md states (ok: false). No stale copy of the superseded reading is left anywhere in the tree.

The --baseline-sample split is unaffected: cli.py:2627-2636 sets payload["ready"] = False and leaves the top-level ok alone, returning 1 through baseline_failed, so changelog.d/1155-baseline-sample-failed-envelope.md's ok: true, ready: false is still the true rendering and is not a second stale line.

Red-count spot-check (same-id comparison, not the full suite): the docker files reproduce 2 failed + 10 errors with identical ids at tip and base; the apply-noop/era-srbench/genesis/skill-library/wheel set reproduces 17 failed with identical ids at both tips; the [tmux] row passes 2/2 at the tip and 1/1 at base. That covers 29 of the 32 red ids; 3 were not sampled.

Blocking: none.

Non-blocking:

  1. The fragment is accurate but narrower than the behaviour: a reader of changelog.d/1093-probe-failed-next.md alone learns that a refused probe is ready: false and nothing about ok. Since both fragments fold into the same release and #1175's carries the ok flip, this is fine as shipped — noted only in case the two fragments are ever separated.
  2. agentdescent plan ... --json exit codes are documented by test, not by --help; nothing in this PR claims otherwise.

Risk / merge: risk_gate → T3; mergeable: true, mergeable_state: blocked; 0 reviews and 0 review comments at this tip, so no PASS names 3b41def3. Advisory only — the counting PASS and the click are the owner's, and a new commit voids this report.

…ment

The refused-probe contract now sets ok:false on both surfaces (#1175), so the pending #1093 fragment must not also claim ok is about the plan itself; both land in the same release. ready:false and the repair action stay.
@github-actions github-actions Bot added the risk:T3 Risk tier T3 (core loop, release, process; needs Danyang's click), set by the controller label Nov 20, 2026
@HugoXK

HugoXK commented Nov 20, 2026

Copy link
Copy Markdown
Collaborator Author

Verdict: PASS (advisory — .github/agent/review-policy.json counts only Birfy)
Tip SHA: a62bcf65424f7546b525b8b98e2b59a13a05687e (PR head matches, non-draft, base 38f9d90f = main, 6 commits, 13 files +143/−33; mergeable_state: blocked)

Ran (this host, clone at the tip, Python 3.10.12 / pytest 9.2.0, PYTHONPATH=., offline)

  • tests/test_plan_probe.py tests/test_mcp_plan_probe.py → 23 passed
  • focused set from the body (test_plan_probe, test_mcp_plan_probe, test_cli_help_epilog, test_cli_help_hides_internal_run, test_call_budget, test_budget_guard, test_integrations, test_plugin_host_cli) → 190 passed, 2 skipped, 1 failed. The failure is test_integrations.py::test_the_launcher_install_writes_ignores_a_shadowing_cwd, and it fails identically on pristine 38f9d90f with the same traceback — mcp_launcher() resolves to /home/xuke/work/agentdescent-reviewer/runs/rev-pr936-8d2fe50/.venv/bin/agentdescent, a console script installed from a different checkout, so this is my review venv, not this PR.
  • red-first at 38f9d90f with only the two test files from this branch → 4 failed, 19 passed, exactly the four ids: KeyError: 'ready' at tests/test_mcp_plan_probe.py:179 and :198; assert True is False at tests/test_plan_probe.py:360 and :251 (the latter carrying a refusal is not success on either surface (#1175)).
  • python -m tools.changelog_fragments check → 1 changelog fragment(s) OK; tools.gen_api_docs --check → docs/api.md is up to date; tools.gen_algorithms_docs --check → docs/algorithms.md is up to date
  • local risk_gate.py --pr 1178 → T3, 176 changed lines, 13 files

Not run

  • full pytest -q; mkdocs build --strict (not installed in my venv); docker rows; wheel smoke
  • CI: the nine checks were not finished at read time — not waited for and not used as evidence

Summary

  • The merge is what it says. fa5fe5b5 has two parents, 3b41def3 and 38f9d90f. The one conflict is the changelog.d/1093-probe-failed-next.md modify/delete; the deletion was taken, the file is absent from the files API list and from changelog.d/, and CHANGELOG.md is untouched by the branch (git diff 38f9d90f..a62bcf65 -- CHANGELOG.md is empty).
  • Nothing was re-verified by luck. git diff 3b41def3..a62bcf65 over agentdescent/cli.py, agentdescent/mcp.py, tests/, tools/, the two SKILL.md copies and integrations/dsh-agentdescent/lib/index.js is empty; the only agentdescent/ change since the sha I judged is main's __init__.py 0.11.0 bump. The exit-1 contract, the ok flip guarded by probe_refused (cli.py:2670), the algorithms run flip and _with_probe's ready-only-when-asked all carry from 3b41def3 byte for byte.
  • The docs merge kept both sides. Pins are main's 0.11.0 (docs/plugins.md:24, docs/plugin-quickstart.md:13, :16) and the probe text sits at docs/plugins.md:258-266 and docs/plugin-quickstart.md:281-284 with no duplicated or dropped sentence; docs/algorithms.md and tools/gen_algorithms_docs.py agree, which the --check run confirms.
  • The quote is exact, not a paraphrase. Normalising whitespace, the fragment's quoted sentence is a substring of CHANGELOG.md at 38f9d90f, (#1093) tail included: “--json sets ready: false while leaving ok about the plan itself, and a successful or omitted probe is unchanged (plan/algorithms run --probe: a failed probe still ends with 'next: agentdescent evolve' (and --json says ok: true) #1093).”
  • 6b034fd2 is real. 6b034fd2c2c022dddb0dc70e63e71993d0e14c2f is main's fix(cli): hide the internal run worker from lab usage (#1172), an ancestor of both 38f9d90f and this tip, three commits behind the release and therefore an older main state, not a branch sha. The body cites it for the first red row and then re-verifies at 38f9d90f; the second row is the decisive one and it is correct. No body edit needed on that account.
  • On the agreement question: no reading survives in which 0.11.0 and 0.11.1 agree. The sentence is framed as what 0.11.0 "said", preceded by "This supersedes the 0.11.0 entry (plan/algorithms run --probe: a failed probe still ends with 'next: agentdescent evolve' (and --json says ok: true) #1093)" and followed by "It now sets ok: false too, on both surfaces", with the unchanged half of the old sentence explicitly still true. The only soft spot is the pronoun in "It now sets": its nearest antecedent is the quoted sentence rather than the behaviour, so a fast reader can momentarily parse it as the old entry continuing — "This PR sets ok: false there too" would remove even that. Optional, not an edit request.

Blocking

None.

Non-blocking

  1. The body's 191 passed is host-specific: on my venv the same eight files give 190 passed, 2 skipped plus the pre-existing launcher failure above. Not a PR defect, but the number is not reproducible as written.
  2. Fragment pronoun noted above (changelog.d/1175-mcp-probe-contract.md:10-13).
  3. Quoting the released entry rather than editing it is the compliant form here (a PR must not touch CHANGELOG.md), so the 0.11.0 text is corrected forward in the next release notes only. Worth the owner seeing once, since 0.11.0 itself stays wrong on disk.

Risk / merge

T3 (176 lines, 13 files) — advisory PASS only. Needs Danyang's PASS at a62bcf65424f7546b525b8b98e2b59a13a05687e and the merge click; a new commit voids this report.

The 0.11.0 release folded changelog.d away, deleting 1093-probe-failed-next.md, which this branch had edited. Take the deletion: the refused-probe contract is stated by 1175-mcp-probe-contract.md and the released CHANGELOG text is corrected there instead. No other path conflicted; docs keep main version pins.
The release shipped the old wording, so the fragment quotes it verbatim before stating the corrected behaviour: a refused probe is ok:false and ready:false on both surfaces.
@Birfy

Birfy commented Nov 20, 2026

Copy link
Copy Markdown
Collaborator

Self-review (Birfy)

Tip SHA: a62bcf6
Ran:

  • .github/scripts/risk_gate.py --pr 1178 → T3 (SKILL.md / dsh plugin paths; 176 lines / 13 files)
  • pytest -q tests/test_plan_probe.py tests/test_mcp_plan_probe.py → 23 passed
  • python -m tools.changelog_fragments check → 1 fragment OK (branch is on main 38f9d90f / v0.11.0)
  • Required CI at this tip → green (build, test 3.9–3.13, otel-wire)
    Not run:
  • Full offline suite locally — the CI test matrix is the line
  • Regenerating the dsh plugin from scratch — the index.js change is the shared SKILL.md body, matching the skill/docs wording
    Findings:
  • No actionable defects. A refused probe now prints ok: false and ready: false on CLI plan / algorithms run and the MCP plan tool; exit code stays 1 (config answer), not 2. The changelog correctly quotes and supersedes the v0.11.0 plan/algorithms run --probe: a failed probe still ends with 'next: agentdescent evolve' (and --json says ok: true) #1093 sentence. The T3 hit on integrations/dsh-agentdescent/lib/index.js is the generated skill string, not a new subprocess path.
    Verdict: PASS

@Birfy
Birfy merged commit 920dc8c into main Nov 20, 2026
10 checks passed
@Birfy
Birfy deleted the agent/1175-probe-contract branch November 20, 2026 10:23
Birfy pushed a commit that referenced this pull request Nov 20, 2026
…`max_calls` / `max_usd` (#1186)

Fixes #1184.

Claimed under AGENTS.md "Claim, PR, release"; agent-ready is triage-side
and applied to 0/18 open issues (repo-wide), so it was not waited on.

## What changed

The run-level call cap (#1166) and the operator ceiling (#1082) were
both unreachable from the MCP surface: `plan` and `start` took
`budget_usd` / `usd_per_call` and nothing else, so a host that wanted
"at most N model calls for this run" had no lever except the
process-wide `AGENTDESCENT_MAX_CALLS`, which it cannot lower per call.

`plan` and `start` (both the tool bodies and the `maybe_tool` wrappers)
now take `max_calls` and `max_usd`. They validate through
`build_call_limit`, the same rule the CLI and the run store use, so a
bad value, or a `max_usd` with no `usd_per_call`, refuses **before a run
directory exists** and with the CLI's own sentence rather than a second
wording for this surface.

The operator's ceiling is applied with the same `min` as
`runstore.launch` / `execute`: a caller may lower it, never raise it,
and a caller that names no cap still gets it. The reply carries
`call_limit`:

- `{"max_calls": 3, "max_usd": null}` — nothing clamped it.
- `{"max_calls": 2, "max_usd": null, "source": "AGENTDESCENT_MAX_CALLS",
"max_calls_requested": 9}` — the ceiling bound a request of 9.
- `{"max_calls": 2, "max_usd": null, "source":
"AGENTDESCENT_MAX_CALLS"}` — the caller named no cap and the ceiling
applies anyway.

`source` and `max_calls_requested` appear only when the ceiling bound
the run, the same distinction `AGENTDESCENT_MAX_CALLS_REQUESTED` already
carries across the detach boundary: without them a host cannot tell "the
ceiling lowered your 9 to 2" from "you asked for 2". A request of
exactly the ceiling reports **not bound**, matching
`_clamped_max_calls`.

`start` passes the caller's own numbers to `runstore.launch`, not a
pre-clamped value: `launch` applies the same `min` and is what records
what was asked for in the child's environment, so handing it an
already-clamped number would make `requested == ceiling` and erase the
label from `stop_detail`, while the arithmetic stayed right. The
enforced clamp still crosses the process boundary as `--max-calls 2`,
which the test asserts on the child's argv. No argument means exactly
today's behaviour, and plain `plan` has no `call_limit` key.

`max_concurrency`, `doctor`, and any USD default for un-priced runs are
untouched, as the issue scopes them out: a process-wide dollar ceiling
would turn every un-priced run into a `SpecError` through
`require_budget_pair`.

## Red test on `f5017f93`

`tests/test_mcp_call_limit.py` alone at clean `f5017f93` → **8 failed, 2
passed**. The eight are the missing entry point, not a behaviour
disagreement:

- `test_plan_reports_the_cap_it_will_enforce` → `TypeError: Tools.plan()
got an unexpected keyword argument 'max_calls'`
- `test_start_caps_the_run_and_status_shows_it`,
`test_max_usd_alone_is_refused_with_the_cli_sentence`,
`test_a_caller_cap_above_the_operator_ceiling_is_clamped_and_labelled`,
`test_a_cap_equal_to_the_ceiling_is_not_reported_as_bound`,
`test_start_hands_the_clamped_cap_to_the_detached_child`,
`test_nested_start_is_unchanged` → the same `TypeError` on `start`
- `test_a_run_that_names_no_cap_still_gets_the_ceiling` → `KeyError:
'call_limit'`: the base payload has no such key, because only the new
path reports the ceiling it will enforce

The two that pass are the two "no cap, behaviour unchanged" rows
(`test_plan_without_a_cap_is_unchanged`,
`test_start_without_a_cap_is_unchanged`), which assert today's behaviour
and must hold before and after.

Provenance: re-measured at the base with only this branch's
`tests/test_mcp_call_limit.py` copied into a clean worktree of
`f5017f93` — `PYTHONPATH=$PWD pytest -o addopts=''
tests/test_mcp_call_limit.py -q` → `8 failed, 2 passed in 2.46s`; all
eight fail on the `TypeError` for the new `max_calls`/`max_usd` kwargs,
and the two passers are the rows named above. The behaviour rows are
asserted green **at this tip**: the run stops at the cap
(`stop_reason="budget"`, the ceiling named in `stop_detail`), the
child's argv carries `--max-calls 2` when the caller asked 9 under a
ceiling of 2, a refused `max_usd` leaves no run behind, and nested
`start` still returns its unchanged stub.

Clamp teeth, measured: with the `min` bypassed in the helper and nothing
else changed, the file went **2 failed, 8 passed**
(`test_a_caller_cap_above_the_operator_ceiling_is_clamped_and_labelled`,
`test_a_run_that_names_no_cap_still_gets_the_ceiling`). The probe was
reverted, not shipped; the file greps clean and is 10 passed.

Tip SHA: 4015bb8

Ran:
- `pytest -o addopts='' -q` (the full offline suite) at `4015bb8f` → 21
failed, 10 errors, 7924 passed, 51 skipped in 2169.68 s. The 31 red ids
are the **identical set** red on `14ba9529`, a commit that is five
commits **older than this PR's base** `f5017f93` — the five being
`38f9d90f` (release v0.11.0), `920dc8c3` (#1178), `98655111` (#1180),
`a1b41e76` (#1181) and `f5017f93` (#1177), i.e. four code merges sit
between the baseline and the base. That is the stronger direction: the
same ids are red *before* those merges as well, so neither this branch
nor the four merges in between accounts for them; the red ids are docker
engine, exec-bit and wheel-smoke environment rows, and the comparison is
a same-id `comm` (empty both ways), not a count. No red id touches
`mcp`, `plan`, the run store caps or this new file. The one count
difference from the earlier `c6000830` measurement is
`test_dsh_superagent_run.py::[tmux]`, which is load-flaky (it passes in
isolation at both commits) and passed in this run.
  - `pytest tests/test_mcp_call_limit.py` → 10 passed
- focused set (`test_mcp_call_limit`, `test_mcp`, `test_mcp_meta`,
`test_mcp_plan_probe`, `test_mcp_http_transport`,
`test_operator_run_cap`, `test_budget_guard`) → 116 passed
- `python -m tools.changelog_fragments check` → `5 changelog fragment(s)
OK`
  - `python -m tools.gen_api_docs --check` → `docs/api.md is up to date`
- `python -m tools.gen_algorithms_docs --check` → `docs/algorithms.md is
up to date`
- `mkdocs build --strict` → rc 0, 12.4 s, with the `social` og-image
plugin stripped in a temporary untracked `mkdocs-local.yml` (it blocks
on a font fetch in this sandbox; CI runs the full config)
Not run:
- real-model probe / USD cap against a live provider → `ARK_API_KEY` is
not in this box's environment; every test row is offline and the cap is
arithmetic, not a model call
  - wheel smoke → no packaging or release path change
  - `doctor`, `max_concurrency` → explicitly out of scope in the issue

No UI surface: this adds two tool arguments, one JSON block and a docs
row; there is no UI to screenshot.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

agent-pr Pull request opened by an agent area:cli The agentdescent command line (cli.py, doctor, plan/evolve/status/apply) area:plugins Host plugins and installers: Claude Code, Codex, DSH, OpenCode, MCP server bug Something isn't working risk:T3 Risk tier T3 (core loop, release, process; needs Danyang's click), set by the controller

Projects

None yet

Development

Successfully merging this pull request may close these issues.

mcp/cli: a refused --probe reports ok:true+ready:false on the CLI but ok:false and no ready over MCP plan

2 participants